Also, the JV-ODE car demo is a quick and simple car set up, it doesn't even remotely represent a completely tuned simulation of a car
That ONLY applies to ODE when compiled with a fixed friction contact mode
I looked into the other friction modes when I wrote the original ODE wrapper ages ago, and I could never get it to work right despite spending about 3 years trying (using it 'properly' with C & OpenGL as well as via the Blitz wrapper) - in theory it seemed you could set 2 different values for friction along the x and z planes, which means you could have a different lateral friction to that which propells the vehicle, which sounds good. Like I said though I never was able to achieve a decent result. Also the rear-wheel steering problem always caused problems at any reasonable speeds.
I'd love to see an ODE demo with the high-speed rotational stability solved, the unwanted rear-wheel steer sorted and the friction allowing for controlled power-sliding - if this was managed and ODE is so much faster than Tokamak as you say then obviously I'd prefer to use ODE (although I'd say that having to call the physics 3 times to alleviate the stability problem negates the speed advantage - speed vs accuracy perhaps?)
Until I see otherwise I think that Tokamak produces cars that are more fun to drive than ODE cars, unless you are willing to delve in Pacejka and some really complex maths, which unfortunately is way beyond me.